Let me get the obvious thing out of the way first. We're an agency. I run one. So writing an article about how to manage agency relationships is a bit like a fox writing a guide to henhouse security. I'm aware of the optics.
But that's actually why I think this is worth writing. I've sat on both sides of every dynamic I'm about to describe. I've seen what clients do that sets projects up to fail. I've seen what agencies do - sometimes what we've done - that creates problems. And I've watched enough engagements go sideways, both ours and other people's, to know that the failure almost never starts where people think it does.
It almost never starts with picking the wrong agency.
We picked a good agency. Surely the relationship will take care of itself.
If you're an operations leader or a CTO at a mid-sized B2B firm, that's probably your default assumption. You did the due diligence. You liked the case studies. The pitch was sharp. The team seemed solid. So you signed, shook hands, and expected things to unfold more or less as described.
And for a while they did. Until they didn't.
I was at dinner with a COO from a financial services firm about six months ago. She'd just come out the other end of a website re-platforming project that went roughly £180k over budget and landed four months late. Good agency, actually - one I know and respect. But she said something that's been rattling around in my head since: "The problem wasn't their capability. The problem was that nobody set the rules of engagement, and by the time we realised we needed them, we were already underwater."
That's the pattern. Nearly every agency relationship failure I've seen in 25 years comes down to one of four things - sometimes a combination, mercifully rarely all four at once.
Unclear scope. And I don't just mean "we didn't have a brief." Most firms have a brief. What they don't have is a shared, unambiguous understanding of what's included, what's not, and what happens when something falls in the grey area between the two. I saw a professional services firm last year that had a 40-page scope document and still ended up in a dispute about whether content migration was included. Forty pages. The word "migration" appeared twice, both times in a context that could be read either way. The agency read it one way. The client read it the other. Nobody checked until month three.
Weak governance. This is the one that kills slowly. No agreed cadence for check-ins, no clear escalation path, no defined decision rights. So when a problem surfaces - and problems always surface - nobody knows who's supposed to deal with it, how quickly, or what authority they have to make trade-offs. The project drifts. Emails pile up. Someone eventually calls a "crisis meeting" that should have been a routine Tuesday catch-up eight weeks earlier.
The wrong commercial model. Fixed price when the scope is genuinely uncertain. Time and materials when there's no cap or accountability mechanism. Milestone payments that don't correspond to meaningful milestones. I've seen firms pay 60% of a project fee before a single line of production code has been written, and then wonder why the agency's urgency seemed to evaporate. The commercial model is the skeleton of the relationship. Get it wrong and everything that hangs off it is misshapen.
The pitch-to-delivery quality gap. This is the one nobody wants to talk about. The people in the pitch aren't the people doing the work. That senior strategist who had you nodding along for 90 minutes? She's on six other accounts. Your day-to-day team is two mid-weight developers and a project manager who joined three weeks ago. I wish I could say this was rare. It isn't.
So how do you actually protect yourself before the contract is signed? The portfolio is the least useful part of the evaluation. Of course it matters - you need to know they've done similar work. But a portfolio tells you what an agency has produced. It tells you almost nothing about what it was like to get there.
If I were hiring an agency tomorrow - and yes, I'm an agency owner telling you how to scrutinise people like me, make of that what you will - here's what I'd actually do.
Talk to the delivery team, not the sales team. Ask to meet the people who'll actually work on your project. If the agency pushes back - "we haven't allocated the team yet" - that's information. It means either the team doesn't exist yet, or they're currently finishing something else that's running late. Neither is necessarily a dealbreaker, but you should know.
Ask for references from projects that went wrong. Any agency can give you three glowing references. I want to hear about the project that hit trouble. How did they handle it? Did they flag it early or hide it? Did they propose solutions or just describe problems? A firm that can give you a reference from a difficult project - and the client still says something positive - that tells you more than ten case studies.
Probe how they handle disagreement. At some point during your project, you're going to disagree about something. Maybe the approach, maybe the timeline, maybe a design decision. Ask the agency how they handle that. If their answer is some version of "the client is always right," be worried. You don't want a yes-agency. You want one that'll push back with evidence and then respect your decision if you still disagree. That dynamic is genuinely rare and genuinely valuable.
Look at process transparency. Can they show you their delivery methodology? Not a glossy PDF about "our agile approach" - the actual tools, boards, and rhythms they use to manage work. At Distinction, we share our project boards with clients from day one. Not because we're unusually virtuous, but because hiding the messy reality of delivery is how trust erodes. If an agency can't show you how they work, they probably don't want you to see it.
Right, so you've picked an agency you feel good about. Now the bit that most firms skip: structuring the engagement so that good intentions actually translate into good outcomes.
Phase the work. Don't sign a single contract for a twelve-month programme. Break it into phases with defined outputs at each stage - discovery and strategy, then design and architecture, then build and launch. Each phase should have its own scope, its own deliverables, and a decision point where you can assess progress before committing to the next stage.
We recommend starting with an assessment before any major engagement. We run these as 14-day sprints - long enough to understand the real problem, challenge the assumptions in the original brief, and produce a recommendation grounded in evidence rather than guesswork. I've seen firms save hundreds of thousands of pounds because a short assessment revealed they were about to commission the wrong project entirely. One COO at a 400-person professional services firm told us: "We were about to spend £400,000 on a project we didn't fully understand. The assessment cost us a fraction of that and told us what we actually needed." What she didn't mention - but I remember clearly - is that the original brief had been sitting in a drawer for eight months while three different internal stakeholders argued about priorities. The assessment settled it in two weeks.
Tie payments to milestones, not time. Monthly retainers are fine for ongoing relationships. For project work, link payments to defined deliverables. Not "end of sprint two" - that's just time with extra steps. "Approved wireframes for all core templates" is a milestone. "Four weeks of design work" is not. The difference matters because it creates a shared incentive: the agency wants to reach the milestone because they get paid, and you want them to reach it because it means tangible progress.
Nail down IP ownership. This one catches people out more often than you'd think. Who owns the code? Who owns the designs? If you part ways, can you take the work and give it to another agency to finish? Get this in the contract - not in an email, not in a verbal agreement. I've seen firms discover mid-project, after a relationship breakdown, that the agency's standard terms give them a licence to use the work but not ownership of it. That's an expensive discovery to make at month seven.
Define exit clauses. Nobody wants to talk about this at the start of a relationship. It feels like writing a prenup on the first date. Do it anyway. What happens if you need to terminate? What's the notice period? What deliverables do you receive for payments already made? If the agency balks at including clear exit terms, ask yourself why.
Here's where I see a lot of firms overcorrect. They've been burned before, so they want to approve every decision, attend every standup, see every wireframe iteration. And in doing so, they slow the project to a crawl and drive the agency's best people to ask for a transfer.
Good governance is about visibility and defined decision rights. That's it.
Weekly status meetings with a fixed agenda. Progress against plan. Blockers. Decisions needed this week. Risks flagged. Thirty minutes. If the meeting regularly runs over an hour, something is structurally wrong with the project, not the meeting.
A clear RACI. Sounds corporate. Don't care. You need to know who's Responsible for doing the work, who's Accountable for decisions, who needs to be Consulted, and who just needs to be Informed. The most common governance failure I see is everyone thinking someone else is making the decision. The RACI doesn't need to be a spreadsheet on a wall - it just needs to exist and be agreed.
Defined escalation paths. When a project manager can't resolve an issue, who does it go to? Two escalation levels is usually enough. The point is that problems don't sit and fester because nobody knows whose job it is to deal with them.
Fortnightly or monthly steering meetings. Separate from the weekly delivery rhythm. This is where you zoom out - are we still heading in the right direction? Has anything changed in the business that should change the project? Are the assumptions we started with still valid? This is the meeting where you catch strategic drift before it becomes a crisis.
Every agency will hit bumps. What's not normal - and what you should treat as genuine warning signs - is a pattern of behaviour that suggests the agency is managing your perception rather than managing the project.
Going quiet between milestones. If you don't hear from the agency for two or three weeks, that's almost never because everything is going smoothly and they're heads-down building. It's usually because they've hit a problem they're hoping to solve before you notice. The best agencies over-communicate, especially when things are difficult. Silence is not confidence.
Scope changes without trade-offs. Your agency comes to you and says, "We've realised we need to add X to the project." Fine - scope evolves, that's reality. But the next sentence should be: "Here's what that means for the timeline and budget, and here are the options." If they present additional scope without acknowledging the impact, they're either not managing the project properly or they're hoping you won't ask. Either way, push back.
The senior team disappearing. Remember that impressive strategist from the pitch? If she hasn't been in a meeting for six weeks, ask about it directly. "We were told Sarah would be involved in this engagement - when was the last time she reviewed the work?" Agencies hate this question, which is exactly why you should ask it.
Blaming your team for delays. Sometimes it is your team causing delays - approvals take too long, stakeholders don't show up to workshops, content isn't provided when promised. Fair enough. But if the agency's default explanation for every missed deadline is "we were waiting on you," start keeping your own records of what was delivered when. It's a pattern that can mask the agency's own delivery problems.
Resistance to sharing working files. If you ask to see the Figma file, the codebase, the project board - and the answer is "we'll share that at the end" or "our process doesn't work that way" - that's a red flag. Transparency isn't a nice-to-have. It's how you verify that the thing you're paying for is actually being built.
One of the things I've learned - sometimes the hard way - is that the best outcomes happen when the agency works with you, not just for you. That's not just a positioning line we use at Distinction because it sounds good on the website. It's a genuine observation about what makes projects succeed or fail.
When we embed alongside client teams, when your people are in the room making decisions with our people rather than receiving a finished thing and reacting to it, the quality goes up and the risk goes down. Your team builds capability. Our team builds understanding. The handover at the end isn't a cliff edge because there was never a clean separation to begin with.
Not every agency should work this way - some engagements genuinely suit a more hands-off model. If you're buying a brochure website, "done for you" is probably fine. If you're re-platforming your digital estate or building something your team needs to own and evolve after launch, "done with you" is almost always the better bet. Worth thinking about which one you actually need before you sign anything.
I know I said I'd keep this as prose, but if you're about to engage an agency, having something you can refer back to is more useful than my opinions about governance theory. So here's what I'd check before signing:
None of this is complicated. That's sort of the point. Engagements fail not because the problems are unpredictable, but because everyone assumes the relationship will handle it, and then it doesn't.
I should be straight about something. Everything I've written here benefits agencies as much as it benefits clients. Clear scope means fewer disputes. Milestone payments mean predictable cash flow. Good governance means problems get caught early instead of festering into something expensive. When a client shows up with a well-structured engagement model, the project is better for everyone.
The firms that structure their engagements properly - clear scope, milestone-based payment, defined governance, actual contractual protections - get dramatically better outcomes than firms that rely on trust alone. Trust matters enormously. But trust without structure is just optimism.
And optimism, in my experience, is not a project management methodology.
If you're about to kick off a major digital programme and you want an honest conversation about how to structure it - whether with us or with someone else - that's exactly the kind of conversation we have in our 14-day assessments. No commitment to a build. Just clarity on what you actually need and how to set it up so it works. You can find out more at distinction.co.uk.